T cell phenotypes associated with metabolically healthy obesity: Results from the Berlin Aging Study

Background In this cross-sectional analysis, we included a sample of 437 older participants (60–84 from the Berlin Aging Study II (BASE-II). Peripheral blood mononuclear cells were isolated, immune cell subsets were analyzed with multiparameter ow cytometry and systemic cytokine levels were measured. Immune cell parameters were correlated with metabolic measures and multiple linear regression analysis was conducted and adjusted for various demographic and clinical factors. in hematologic parameters, and Plasma C-reactive protein levels All hematologic parameters and CRP normal BMI 31.8% 28.8% dened OGTT OGTT HOMA-IR


Abstract Background
Obesity is associated with chronic low-grade in ammation leading to metabolic and cardiovascular diseases, but a subset of obese individuals is considered metabolically healthy. The underlying pathophysiologic mechanisms remain elusive and clinical studies on the relationship between in ammatory markers and metabolically healthy obesity (MHO) are scarce.

Methods
In this cross-sectional analysis, we included a sample of 437 older participants (60-84 years) from the Berlin Aging Study II (BASE-II). Peripheral blood mononuclear cells were isolated, immune cell subsets were analyzed with multiparameter ow cytometry and systemic cytokine levels were measured. Immune cell parameters were correlated with metabolic measures and multiple linear regression analysis was conducted and adjusted for various demographic and clinical factors.

Results
We found that frequencies of naïve and memory CD4 + and CD8 + T cells inversely correlated with measures for insulin sensitivity in the older population. Moreover, the percentages of naïve CD4 + and CD8 + T cells were signi cantly higher, whereas activated T cells and IL-6 levels were lower in metabolically healthy compared to metabolically unhealthy obese individuals. The percentages of naïve CD4 + and CD8 + T cells were predictive for impaired insulin sensitivity (ß=0. 16, p = 0.01 and ß=0.11, p = 0.04), and the association of naïve CD4 + T cells with insulin sensitivity persisted after multivariate adjustment (ß=0.14, p = 0.02).

Conclusions
These ndings support the hypothesis that parameters of systemic in ammation can differentiate metabolically healthy from metabolically unhealthy obese individuals that are at higher risk for cardiometabolic diseases and may have clinical implications with regard to obesity treatment strati cation.
Trial registration DRKS00009277. Registered 31 August 2015 -Retrospectively registered, https://www.base2.mpg.de/de Background Obesity, de ned by the accumulation of abnormal or excess body fat, is a worldwide epidemic, with increasing prevalence in developed and developing countries and with a serious impact on human health.
Worldwide, over 2 billion people are overweight or obese (1). Obesity is a major risk factor for metabolic diseases, such as type 2 diabetes, dyslipidemia, fatty liver disease and cardiovascular diseases, such as hypertension, peripheral artery disease, myocardial infarction, stroke, and cancer. (2)(3)(4)(5). However, the large variation in the individual risk to develop insulin resistance or other obesity-related comorbidities has led to the concept of metabolically healthy obesity (MHO) (6). Whereas the majority of obese people develop insulin resistance, about 20% are considered metabolically healthy with an age-and genderdependent prevalence, and, thus, seem to be protected from insulin resistance (7). The diagnosis of obesity and treatment options are based on the body mass index (BMI), which fails to reliably predict the predisposition to cardiometabolic diseases (8,9). Individuals with MHO have been de ned by the absence of metabolic impairments or cardiovascular diseases, due to the lack of standardized MHO criteria. However, most recently the BioShare-EU project has proposed a uni ed de nition of MHO, comparing clinical and metabolic factors of 10 population-based cohort studies from 7 countries, including fasting blood glucose (10,11). To guide personalized and risk-strati ed obesity treatment, there is a critical need to investigate differences between MHO and metabolically unhealthy obesity, but informative studies are, thus far, very limited (7). Obesity is associated with systemic and adipose tissue in ammation (12)(13)(14). Excessive energy intake is thought to be a major contributor of obesity, and leads to the accumulation of lipids in adipocytes and the expansion of adipose tissue. Hypertrophic adipocytes can produce pro-in ammatory cytokines, such as IL-6 and TNF-α, which leads to increased vascular permeability, circulating monocytes and the recruitment of chemokines, initiating an in ammatory process that promotes insulin resistance (14,15). Recently, it has been reported that increased systemic and adipose tissue in ammation can differentiate obese women with impaired glucose tolerance from obese women with normal glucose tolerance (16). Blüher and colleagues have shown that macrophage in ltration into visceral adipose tissue, higher in ammatory parameters and circulating adipokines can predict insulin resistance in morbidly obese patients (17). However, only limited and partially discordant data are available on the association of systemic and adipose tissue in ammation with MHO, especially in older patients.
In the present study, we focused on identifying relationships between the distribution of peripheral immune cell subsets, cytokine levels, and metabolic measures in a large non-clinical sample of older adults. Using data from BASE-II, we tested the hypothesis that individuals with obesity and insulin resistance have increased pro-in ammatory cytokines correlating with activated T cell subsets. The aim was to compare a cohort with MHO (obese participants without insulin resistance), a cohort of obese participants with insulin resistance, a cohort of non-obese participants without insulin resistance, and a cohort of non-obese participants with insulin resistance.

Characteristics of the sample
We analyzed 437 of 1671 older BASE-II participants (60-84 years) of BASE-II, medically assessed at baseline, based on the availability of immunological parameters. The sample analyzed in this study was composed of participants with an equal gender distribution. Table 1 shows the median age, BMI, Page 4/21 metabolic measurements (HbA1c, insulin sensitivity index (ISI OGTT ), homeostasis model of assessment for insulin resistance (HOMA-IR)), glucose and insulin parameters measured in oral glucose tolerance tests (OGTT) and lipid parameters), hematologic parameters, and Plasma C-reactive protein (CRP) levels of male and female participants. All hematologic parameters and CRP levels were within the normal range (NR). Obesity was noted in 28 of 181 men (15.5%) and 49 of 256 women (19.1%) (de ned as BMI > 30 kg/m 2 ), whereas 6.6% (women) and 11% (men) were diagnosed with type 2 diabetes. However, 31.8% of the men and 28.8% of the women were considered insulin resistant (IR) at least to some extent, with insulin resistance being de ned as ISI OGTT < 4. Surprisingly, only 50% of all obese participants analyzed in this study had an ISI OGTT < 4, and only 32% of all obese participants were considered truly IR, as assessed by HOMA-IR (HOMA-IR > 2.9). Impaired insulin sensitivity is associated with increased T cell senescence To assess the association of metabolic parameters (BMI, HOMA-IR, ISI OGTT ) with systemic leukocyte subpopulations, we analyzed multiparameter ow cytometric data derived from blood samples of older participants. As expected, the BMI was negatively associated with ISI OGTT (Fig. 1A). The ISI OGTT also correlated negatively with the total number of leukocytes ( Fig. 1B), in line with previous results from another study of ours (18). Whereas the percentages of major mononuclear leukocyte subsets (CD4 + and CD8 + T cells, B cells, NK cells, and monocytes) did not signi cantly correlate with BMI, HOMA-IR or ISI OGTT , we found that the percentages of naïve (CD45RA + CCR7 + ) CD4 + and CD8 + T cell subsets were positively associated with ISI OGTT and negatively associated with HOMA-IR in older participants, whereas the percentage of central memory (CD45RA − CCR7 + ) CD4 + and CD8 + T cells was negatively associated with ISI OGTT and positively associated with HOMA-IR ( Fig. 1C FoxP3 + Treg cells, which have been reported recently to be critical in prevention of autoimmune-mediated diabetes, correlated positively with BMI (data not shown). Additionally, we correlated cytokine levels (IL-6, IL-10, TNF-α, IL-1ß) with metabolic measures. As described previously (20), the seminal in ammatory marker IL-6 was negatively associated with ISI OGTT (Fig. 1H), but the associations between IL-10, TNF-α, IL-1ß and metabolic measures were not signi cant (data not shown). CD57 has been proposed as a marker for T cell senescence, as its expression is associated with impaired proliferative capacity and other characteristics of senescence (21). Here, we found a positive association of CD57 expression on central and memory CD8 + T cells with BMI and HOMA-IR, whereas CD57 + central memory CD8 + T cells correlated negatively with ISI OGTT (Supplementary Table 1). With age, the distribution of circulating T cells at different stages of differentiation changes drastically, as a result of the minimal production of naïve T cells (CD45RA + CCR7 + ), and the accumulation of antigen-experienced memory cells, some being highly differentiated senescent-like T cells (CD45RA − CCR7 − PD-1 + ). This could contribute to increased severity of infections in the older population (22). Taken together, our ndings thus imply that insulin resistance is associated with a higher "age" of circulating T cells.
T cells have a more activated phenotype in metabolically healthy versus unhealthy obese individuals To better characterize the immunological differences between insulin sensitive (IS), who are considered metabolically healthy, and IR (considered metabolically unhealthy) obese and non-obese participants of this study, we strati ed the cohort into four groups by the ISI OGTT . The frequencies of naïve CD8 + T cells were signi cantly higher in obese and non-obese IS compared to IR participants, whereas the frequencies of effector memory and CD57 + antigen-experienced and differentiated CD8 + T cells were signi cantly lower in obese IS compared to IR individuals ( Fig. 2A-C). Additionally, IL-6 levels, which have previously been reported to promote insulin resistance (20), were higher in obese IR than IS individuals (Fig. 2D). In the CD4 + T cell compartment, frequencies of naïve CD4 + T cells were higher and frequencies of central memory CD4 + T cells were lower in IS versus IR non-obese individuals, while we did not see any signi cant differences in the obese group ( Fig. 2E-G, Supplementary  Table 2). In addition, the levels of the anti-in ammatory cytokine IL-10 ( Fig. 2H), TNF-α and IL-1ß were not signi cantly different in the obese and non-obese subgroups (Supplementary Table 3). TNF-α and IL-1ß were undetectable in some participants, and these were excluded in further studies. These ndings suggest that senescence of circulating T cells re ects a major difference between IS and IR participants in both obese and non-obese subgroups.
The percentage of naïve CD4 + and CD8 + T cells predicts impaired insulin sensitivity Next, we tested the association of ISI OGTT with systemic leukocyte subsets and cytokine levels in multivariable linear regression models. The frequencies of T cells, Treg cells, monocytes, B cells, and NK cells were not signi cantly associated with ISI OGTT , neither after adjustment for sex and BMI (model 1) nor with additional adjustment for the morbidity index and CMV status (model 2). More details of these results that are in line with our ndings that the proportions of major innate and adaptive leukocyte subsets (T cells, Treg cells, monocytes, B cells, and NK cells) were not signi cantly different in obese or non-obese IR and IS participants can be found in Supplementary Table 4. T stem cell-like memory T cells (TSCM, de ned as CD45RA + CCR7 + CD95 + ) were also not signi cantly associated with ISI OGTT (Supplementary Table 4).  Table 3). We replicated these ndings using an additional binary logistic regression model (Supplementary Table 5). To account for potential selectivity based on high age, we also added a sensitivity analysis, and ruled out participants, that were aged 80 years or older (8 participants, age 80-84 years). None of those participants were obese and the results remained unchanged (Supplementary Table 6). CMV was included in this analysis because latent infection with this herpesvirus strongly in uences circulating naïve and memory T cell phenotypes (23). Circulating IL-6 and IL-10 levels were not signi cantly associated with ISI OGTT (data not shown). Taken together, these data indicate that the differentiation and activation state of CD4 + and CD8 + T cells is strongly associated with insulin resistance.

Discussion
The prevalence of obesity has dramatically increased in all age groups in recent years, but obesity rates among older adults are even higher. Increasing age has been shown to be associated with lower prevalence of MHO (10). Recently, it has been reported that obesity-related comorbidities and conditions mirror those of aging and age-related diseases (24). Obesity and aging can lead to chronic low-grade in ammation and an increased incidence of chronic in ammatory diseases due to dysregulated immune responses (25)(26)(27)(28)(29). Macrophages seem to be primarily involved in obesity-associated in ammation, changing their phenotype from "alternatively activated" to "in ammatory" macrophages (30,31). Moreover, other components of the innate immune system, mast cells, neutrophils, and dendritic cells, have been shown to exacerbate insulin resistance (32)(33)(34), whereas eosinophils and type 2 innate lymphoid cells can protect against adipose tissue and islet in ammation (33). More recent work focused on adaptive immune responses in obesity-induced systemic and adipose tissue in ammation. T cells, including CD8 + T cells, Th1, Th17, as well as B cells, can exacerbate in ammation, whereas Treg cells and Th2 cells can dampen in ammation and protect against insulin resistance (35)(36)(37). Several human studies investigated the association of in ammatory parameters with impaired insulin sensitivity. In particular, it has been found that the helper T cell composition in peripheral blood correlates signi cantly with the HOMA-IR (38) and other measures of adiposity, in ammation and glucose intolerance, whereas circulating Treg cells were reduced in obese subjects, and might identify individuals at increased risk for cardiovascular comorbidities (39,40). Moreover, a distinct phenotype of Treg cells has been characterized in human obese omental adipose tissue (41). Another recent study revealed an impaired NK cell phenotype and NK cell subset alterations in obese individuals (42). Reduced circulating Treg cell numbers were detected in obese compared with non-obese study participants (39), and another study identi ed a signi cant inverse correlation of Th2 cells in peripheral blood with systemic insulin resistance (12).
Additionally, our group found recently that insulin resistance correlates signi cantly with a shift in the ratio of naïve and differentiated memory CD4 + and CD8 + T cells in abdominal subcutaneous adipose tissue in female obese subjects (18). Furthermore, it has been found that peripheral frequencies of Thelper (Th)22 cells and IL-22 levels were increased in obese subjects with or without type 2 diabetes compared with lean subjects, and that Th22 cell frequencies correlated positively with HOMA-IR (38).
These ndings were con rmed in another study, showing that Th22 and Th17 cells were elevated in abdominal subcutaneous adipose tissue from metabolically abnormal IR obese compared with metabolically normal IS obese subjects (43). Although the association of obesity with systemic low-grade in ammation is well established, studies on the characteristics of metabolically healthy versus metabolically unhealthy obese individuals are limited.
Here, we have investigated the association of insulin resistance and immunological parameters in a sample of 437 older participants of the BASE II study. We analyzed peripheral blood immune cell subsets and cytokine levels with multiparameter ow cytometry analysis. We found that frequencies of naïve CD4 + and CD8 + T cells correlated positively with ISI OGTT , but negatively with BMI and HOMA-IR, whereas frequencies of central memory CD4 + T cells correlated negatively with ISI OGTT , but positively with BMI and HOMA-IR. Additionally, the percentage of highly differentiated effector memory CD8 + T cells was positively associated with HOMA-IR, and the expression of CD57, a surface marker putatively associated with impaired proliferation capacity and cell senescence (21), correlated positively with BMI and HOMA-IR in older participants of this study. However, the expression of PD-1, a characteristic marker for T cell exhaustion, on CD8 + T cells was not associated with metabolic measures (Supplementary Table 1). In line with previous reports (20,44), we also identi ed a positive association of systemic IL-6 levels with insulin resistance, whereas other cytokines (IL-1ß, TNFα, IL-10) did not correlate signi cantly (Fig. 1). To address the association of MHO with insulin resistance, we divided the study participants in obese and non-obese IR and IS subgroups, based on ISI OGTT . The increased frequencies of peripheral blood CD4 + T cells in IS obese and non-obese individuals were accompanied by a selective increase of naïve CD4 + T cells. Similarly, in the CD8 + T cell compartment, the frequencies of naïve T cells were higher in IS obese and non-obese subgroups, whereas effector memory T cells were signi cantly lower. Altogether, the CD4 + and CD8 + T cell compartment was skewed towards a memory T cell phenotype in IR subjects (Fig. 2, Supplementary Table 2-3). Additionally, the frequencies of naïve CD4 + and CD8 + T cells were predictive for ISI OGTT , and the relationship of ISI OGTT with naïve CD4 + T cells remained signi cant after adjustment for sex, BMI, clinical conditions (morbidity index) and CMV-serostatus (Table 2).
Here, we de ned parameters for insulin sensitivity to de ne MHO, but more recently, the following criteria have been proposed in addition to the diagnosis of obesity (BMI > 30 kg/m 2 ): fasted serum triglycerides ≤ 1.7 mmol/l (≤ 150 mg/dl); HDL cholesterol serum concentrations > 1.0 (> 40 mg/dl) (in men) or > 1.3 mmol/l (> 50 mg/dl) (in women); systolic blood pressure (SBP) ≤ 130 mmHg; diastolic blood pressure ≤ 85 mmHg; fasting blood glucose ≤ 6.1 mmol/l (≤ 100 mg/dl); no drug treatment for dyslipidemia, diabetes, or hypertension; and no cardiovascular disease manifestation (7). Interestingly, frequencies of systemic naïve CD4 + and CD8 + T cells of older participants of the BASE-II study, were positively associated with HDL cholesterol serum concentrations, whereas associations with fasting blood glucose and fasted serum triglycerides were negative (data not shown). However, we did not assess these criteria to de ne MHO in the present study, as most of these criteria are affected by age and can also in uence each other.
Our study has several important limitations. First, we could only include a small subgroup of BASE-II study participants based on the availability of ow cytometric data and cytokine level measurements of blood samples. Although the distribution of the analyzed subgroup is similar to the whole cohort (e.g. equal gender distribution), the sample size is rather small after further subdivision into obese and nonobese, IS and IR groups, and the different numbers of IS (n = 243) versus IR (n = 84) non-obese participants could limit the power to detect signi cant associations. However, the sample size of obese IS (n = 28) and IR (n = 32) participants is very similar, and with regard to the multiparameter ow cytometry analysis we conducted, the sample size is still reasonable. Moreover, the relative homogeneity of the participants with regard to age (65-80 years) could strengthen our study; age-related changes in immune function and T cell alterations have been described previously (23,26).
Second, we measured insulin sensitivity using HOMA-IR and ISI OGTT obtained from OGTT, whereas the hyperinsulinemic-euglycemic clamp technique is considered the most reliable method available for estimating insulin resistance and is used as reference standard. In the present study, only half of the obese subgroup was considered IR with ISI OGTT < 4 and 32% of the obese subgroup was considered IR with HOMA-IR < 2.9. On the other hand, 19% of the non-obese subgroup was considered IR with ISI OGTT < 4, and 4.6% with HOMA-IR > 2.9. Due to signi cant inter laboratory variations in insulin assays, the normal range of these parameters needs to be established for each laboratory, and could also explain differences in the percentages of participants considered IR, here. However, the measurements of HOMA-IR and ISI OGTT are minimally invasive, and, thus, still suitable for clinical uses. Moreover, these surrogate parameters of insulin resistance are widely used in observational studies which allows for comparison between different studies.
Third, we assessed peripheral blood immune cell frequencies in this study, which often correlate with immune cell pro les in adipose tissues (12), but further investigation on immune cell parameters in abdominal subcutaneous and visceral adipose tissue and adipose tissue dysfunction (45) is needed to elucidate biological mechanisms linking obesity to insulin resistance.

Conclusions
In conclusion, our study underscores the role of immunological parameters in the differentiation of metabolically healthy from IR individuals with obesity. Our results suggest that the peripheral blood T cell compartment of IS individuals is characterized by higher frequencies of naïve CD4 + and CD8 + T cells, whereas differentiated and activated memory CD4 + and CD8 + T cell frequencies are lower than in IR individuals. Further studies are needed to explore mechanisms underlying the relationship between T cell senescence and insulin resistance in obesity, and to characterize immunological parameters of MHO in order to guide risk-strati ed obesity treatment.

Study population
We analyzed a subgroup of 437 participants of the BASE-II study selected on the basis of the availability of immune cell parameters, randomly selected from all 1600 participants (Table 1). BASE-II is a prospective multidisciplinary and multi-institutional study that investigates factors associated with aging trajectories in Berlin (46)(47)(48). Phenotypic assessments include factors related to geriatrics and internal medicine, immunology, genetics, psychology, sociology and economics. Initial medical assessments included 2171 participants (∼75% aged 60-84 years and ∼25% aged 20-35 years) (49). Based on the BMI and the ISI OGTT , we divided the participants into four groups: obese and non-obese (BMI > 30 kg/m 2 and BMI < 30 kg/m 2 ) and IR (ISI OGTT < 4) and IS (ISI OGTT > 4). All participants gave written informed consent to the study protocol which was approved by the Ethics Committee of the Charité-Universitätsmedizin Berlin (number of the ethical approval: EA2/ 029/09).

Biochemical measurements
A peripheral venous blood sample of all BASE-II participants was drawn in the morning after an overnight fast (> 8 hours) and kept at 4° C until analysis on the same day. Serum concentrations of total cholesterol, low-density lipoprotein, cholesterol, high-density lipoprotein cholesterol, and triglycerides were measured using enzymatic colorimetric tests or photometric measurements. Glucose levels (fasting and 2-hours post load) were measured using photometric methods and insulin levels were determined by an electrochemiluminescence immunoassay (Elecsys® Insulin, Cobas/Roche). HbA1c was measured using high-performance chromatography (VARIANT II TURBO HbA1c Kit -2.0, Bio-Rad). Insulin resistance was determined using the HOMA-IR (calculated as the product of fasting glucose and fasting insulin divided by 22.5). Moreover, insulin sensitivity was estimated by calculating ISI OGTT based on the work of Matsuda and colleagues (50). CRP levels were determined using an immunological turbidity assay (cobas/Roche, Rotkreuz, Switzerland), as described previously (51).

Additional data
Body weight was measured with a portable electronic scale to the nearest 0.1 kg and height was determined to the nearest 0.1 cm by using an electronic weighing and measuring station (seca 764, seca, Hamburg, Germany). Weight and height were used for calculating the BMI, kg/m 2 . Morbidity was assessed as a morbidity index based on most of the categories of the comorbidity index originally described by Charlson and collaborators (52), which is a weighted sum of moderate to severe, mostly chronic illnesses, including cancer (e.g., lymphoma) and cardiovascular (e.g., congestive heart failure) and metabolic diseases (e.g., diabetes mellitus). The morbidity index used in BASE-II has been described previously in detail (53).
Flow cytometry on peripheral blood mononuclear cells Venous blood was taken from the participants of the BASE-II study during medical examinations at the Charité in Berlin and sent to the BASE-II partner site at the University Tübingen in EDTA tubes (7 ml) packed in iso-containers, to minimize temperature variations. The peripheral blood mononuclear cells (PBMC) were further isolated under sterile conditions and frozen at -196° C in liquid nitrogen until further processing. Flow cytometry surface staining was performed as described previously (54). Treg cells were de ned as FoxP3 + CD25 high cells within the CD4 + or CD8 + T cell subset, as described previously (56). The gating strategy for T cells is shown in Supplementary Fig. 1 and Supplementary   Fig. 2. The gating strategy for B cells has been reported elsewhere (54). Flow cytometry staining and data analysis were performed on blinded samples.

Cytokine analysis
Serum cytokine levels were determined as described previously (29,57). Brie y, cytokine levels of IL-1ß, IL-6, IL-10 and TNFα were determined using the high sensitivity CBA ex system (BD Biosciences), according to the manufacturer´s instructions. Samples were measured (BD LSR-II) and consistent performance was assured using BD CS&T beads.

CMV Serology
Anti-CMV IgG titres were analyzed using a CMV IgG Kit (Omega Diagnostics Group, Scotland, UK) as described previously (51). IgG levels were measured using a semi-quantitative approach, according to the manufacturer's instructions.

Statistical analysis
The results are shown as median 25th, and 75th percentile for continuous variables or as absolute numbers and percentages for categorical variables unless otherwise noted. All statistical analyses were performed with IBM SPSS Statistics software package, version 25 and GraphPad Prism version 8 (GraphPad Software, San Diego, CA). A one-sample Kolmogorov-Smirnov test was used to test variables for Gaussian distribution. The spearman rank correlation coe cient was used for analyzing bivariate correlations between the ISI OGTT and immune cell subsets and cytokine levels. The Mann-Whitney U test was applied to estimate differences between groups. Associations of systemic immune cell subsets and cytokine levels with metabolic measures were analyzed by linear regression models adjusted for sex, BMI, CMV-serostatus and the morbidity index. Because data analyses were exploratory no adjustment was made for multiple testing and p values were interpreted descriptively. An acceptable level of statistical signi cance was established a priori at p < 0.05.  Data are shown as mean ± SEM.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.